ai.bythebay.io: Josh Hartung Interview
FunctionalTV interview with Josh Hartung.
Follow the words
Read the transcript
you you [Music] you boy my definition of a is so we work boy my definition of a is so we work mostly in AI enablement and so well not mostly in AI enablement and so well not an AI enablement but Morris of an AI enablement but Morris of autonomous systems in enablement so one autonomous systems in enablement so one component of that of course is AI so I component of that of course is AI so I think that AI is a general computing think that AI is a general computing engine it's a one of the best engine it's a one of the best definitions that I've sort of scene I definitions that I've sort of scene I was reading a research paper in fact was reading a research paper in fact from Nvidia and they sort of from Nvidia and they sort of characterized an AI model as as a characterized an AI model as as a generalized way to model some type of generalized way to model some type of system and I think that's a really nice system and I think that's a really nice way to say it so you have some type of way to say it so you have some type of input and some type of output and some input and some type of output and some function in the middle that you know can function in the middle that you know can be defined you know in more classical be defined you know in more classical approaches mathematically and in more approaches mathematically and in more you know complex AI based approaches you know complex AI based approaches also mathematically but in much more also mathematically but in much more complex terms AI has been very very complex terms AI has been very very successful in in characterizing successful in in characterizing uncertain or unstructured type uncertain or unstructured type environments so from perception environments so from perception perspective from a planning perspective perspective from a planning perspective it's it's incredibly effective at taking it's it's incredibly effective at taking things that using generalized things that using generalized representations to understand the world representations to understand the world around it and to and to make plans for around it and to and to make plans for achieving goals that's been that's been achieving goals that's been that's been incredible revelation in autonomous incredible revelation in autonomous driving where suddenly you move from driving where suddenly you move from from a system which has to be defined from a system which has to be defined based on rules and you know and maybe based on rules and you know and maybe sort of simple learning behaviors to sort of simple learning behaviors to something which is which exhibits something which is which exhibits extremely complex behaviors with a extremely complex behaviors with a relatively simple sort of development relatively simple sort of development process one of the challenges of AI is that it's one of the challenges of AI is that it's non deterministic I mean in effect it is non deterministic I mean in effect it is deterministic but the input space is so deterministic but the input space is so large that it's difficult to to tie down large that it's difficult to to tie down in terms of a deterministic result for in terms of a deterministic result for any given input and so the challenge any given input and so the challenge there is that is that typically safety there is that is that typically safety models just in a general way rely on models just in a general way rely on determinism the idea that you know we determinism the idea that you know we can we can reliably tie some input to can we can reliably tie some input to some known output and so with AI we some known output and so with AI we really end up violating that principle really end up violating that principle and so I think the most interesting area and so I think the most interesting area of AI from our perspective and we are an of AI from our perspective and we are an infrastructure and safety type company infrastructure and safety type company is how do we create a framework that is how do we create a framework that allows these these I I algorithms to allows these these I I algorithms to perform there you know at their best but perform there you know at their best but also ensuring that they're they're also ensuring that they're they're behaving safely safely I think the body behaving safely safely I think the body of a conference I wasn't sure what to of a conference I wasn't sure what to expect my first time here and I was expect my first time here and I was really impressed it generally at the really impressed it generally at the audience typically I'm talking to audience typically I'm talking to autonomous driving people and we get autonomous driving people and we get much more into the into the weeds in much more into the into the weeds in terms of you know the specifics of of terms of you know the specifics of of our technology or challenges that we've our technology or challenges that we've that we've identified but but talking to that we've identified but but talking to talking to the people here they are a talking to the people here they are a bunch of very very smart people with bunch of very very smart people with lots of experience in in AI but not much lots of experience in in AI but not much experience in an autonomous driving i'm experience in an autonomous driving i'm seeing like a really broad cross-section seeing like a really broad cross-section of folks that are you know working in of folks that are you know working in data science and that I think is just data science and that I think is just great so the audience has been really great so the audience has been really amazing also the speaker line-up I think amazing also the speaker line-up I think it's been really dynamic you know very it's been really dynamic you know very different perspectives different areas different perspectives different areas of interest for each for each speaker so of interest for each for each speaker so those things been great and then I think those things been great and then I think finally the it just written been really finally the it just written been really professionally run like you guys are professionally run like you guys are very it's just very dialed in terms of very it's just very dialed in terms of all the logistics all the all the all the logistics all the all the facilities and things like that so yeah facilities and things like that so yeah just well done I'm impressed just well done I'm impressed you you [Music] you you
Recovered English captions. Automatic transcription may contain errors.
Keep exploring
Follow the guest, their work, and the ideas behind this conversation in the Devreal knowledge graph.
Josh Hartung on Devreal ↗